On Jan 13, 2012, at 8:03 AM, Robert Haas wrote:
> Or,
> we could try to lift this restriction in the special case when
> hot_standby_feedback is set, though I have a feeling that's not really
> robust - any time you lose the connection to the master, it'll lose
> your xmin holdback and possibly mark some things all-visible that
> really aren't on the standby, and then you reconnect and replay those
> WAL records and bad things happen.
Also, what happens if you started off with hot_standby_feedback turned off? New stuff won't just magically start
workingwhen you turn it on (or is that parameter settable only on restart?)
ISTM that hot standbys need some ability to store some "interim data" that is only needed by the hot standby while
oldertransactions are running. IIRC we've seen other places where we have this problem too.
Perhaps it would be possible to keep older copies of pages around when there are older transactions running on the
standby?
--
Jim C. Nasby, Database Architect jim@nasby.net
512.569.9461 (cell) http://jim.nasby.net
